Fujitsu | March 31, 2021
Products are routinely examined for abnormalities, flaws, or manufacturing errors as part of the production process. Now, modern AI technology developed by the Japanese company Fujitsu has the potential to significantly reduce the number of working hours needed for this procedure.
Fujitsu Laboratories, a research center within the multinational ICT group, has developed an AI technology for image inspection that helps for the highly detailed identification of a wide range of external defects on manufactured goods.
The technology employs an AI model trained on photographs of goods with artificial abnormalities, eliminating the need to prepare training data from live images of faulty products pulled from a manufacturing line's inspection procedure.
The problem in training AI to execute quality management tasks is that models are often trained using weighted and summed indices for individual characteristics. This will make it difficult to construct a model that thoroughly comprehends all of the features that must be examined.
Fujitsu has created a system for training an AI model that allows a normal image with several irregularities such as form, scale, and color to be recovered by artificially applying the virtual defects to a normal image prepared for training.
According to the company, this technology achieved more than 98 percent in an AUROC ranking, which is an assessment metric used to evaluate classification model performance.
The score was obtained in a category of items of typical appearance differences, such as carpets with varying fur patterns and colors on an individual basis, and printed circuit boards with different wiring shapes on different components.
Fujitsu has validated the technology's usefulness in a real-world environment at Fujitsu Interconnect Technologies' Nagano factory, which produces electronic equipment. The AI, according to the firm, reduced the number of hours needed for inspecting printed circuit boards by 25%.
In recent years, the manufacturing industry has been a significant target for digitalization. An €11 million European initiative to use machine learning to improve processes and achieve zero-defect manufacturing was announced last week.
Fujitsu's AI advances have also been applied in other fields. Earlier this month, the company unveiled specifics of a facial recognition system that uses an AI model to track subtle variations in muscle expressions in a person's facial expression to measure how focused someone is in online classes and meetings.
IndustryWeek | May 11, 2020
Even in an environment where remote work is normalizing and investments in robotics are paying rewards, robots are still not taking over the production environment. In many of the best use cases robots remain complementary – to humans as well as other technologies.
The human element is critical to making AI and robots work, explains Daniel Theobald, CEO of Vecna Robotics. “As counterintuitive as it sounds, human workers are essential to all manufacturing processes that include autonomous systems. In a facility that employs AI, robots and workers, each plays a unique role,” he says.
AI is constantly scanning, looking for and learning from opportunities to create greater operational efficiencies, and communicating with workers through wearables and robots to deploy them to solve problems in real-time. Robots are used for automating routine, repetitive, dangerous jobs, such as lineside replenishment, racking, and staging. Supported by AI and automation, humans can then turn their attention to higher-value activities, like exception handling and creative problem-solving.
Mech-Mind Robotics | June 09, 2021
Mech-Mind Robotics, a fast-growing Chinese AI startup, has announced its new-generation Mech-Eye Pro Enhanced Industrial 3D Camera. Equipped with fast, structured light, this 3D camera presents high-quality 3D imaging of a broad range of objects, including tiny, shiny, reflective, and colored objects.
The industrial 3D camera has become an essential partner for pick-and-place robotics. Yet, when it comes to densely packed, randomly placed, and highly detailed objects, limitations in resolution, precision, and speed restrict their performance and potential.
With advances in resolution and precision, Mech-Eye Pro Enhanced 3D industrial camera can generate complete, detailed, and accurate point cloud with native and perfectly aligned color for a wide variety of objects, including plastic, ceramic, metal, cardboard, and wood.
In addition, Mech-Eye Pro Enhanced's notable features include:
• Extra field-of-view with 3D imaging time shortened to 0.5 seconds possible. This enables Mech-Eye Pro Enhanced to meet various typical applications, including bin picking, inspecting, measuring, and academic research.
• Compact Aluminium housing with flexible mounting options. This minimizes the impact on maneuverability and enables robot cells to maintain reliable operation in demanding applications.
• Cost efficiency. Mech-Eye Pro Enhanced is competitive in terms of pricing. Moreover, it helps customers boost efficiency and productivity.
• Industrial suitability with IP65, vibration resistant, EMC/EMI proved, obtaining CE, FCC, VCCI, and RoHS certifications.
Mech-Mind has developed an entire infrastructure and industrial 3D cameras portfolio, including:
• Mech-Eye Laser: Equipped with fast, structured light, Mech-Eye Laser boasts ambient light resistance, significantly reducing shading facility requirements even under sunlight exceeds 10000 lux.
• Mech-Eye Nano: Combining small size with high-quality imaging processing, Mech-Eye Nano helps robotic arms better understand and interact with their surroundings.
• Mech-Eye Deep: With a large field of view and extra depth accuracy, Mech-Eye Deep can well recognize tightly-packed wrinkled sacks of different sizes.
With a multi-language SDK for platforms, end-users can integrate Mech-Eye 3D Camera with Mech-Vision graphical vision software and Mech-Viz intelligent robot programming environment to develop customized applications.
With such features and potential, Mech-Eye 3D industrial cameras have been deployed in over 1000 applications across wide-ranging industries, including manufacturing, steel, food, logistics, finance, and healthcare, since 2016.
About Mech-Mind Robotics
Mech-Mind was founded in 2016 to put intelligence into industrial robots. Integrating advanced technologies including deep learning, 3D Vision, and motion planning, Mech-Mind offers cost-effective solutions to palletizing and depalletizing, bin picking, order sorting, machine tending, and assembly/gluing/locating in logistics and manufacturing.
Mech-Mind's intelligent industrial robot solutions are used in automotive OEM plants, appliance plants, steel plants, food plants, logistics warehouses, banks, and hospitals in China, Japan, South Korea, Germany, and the U.S.